Intel announces 48 core Cascade Lake Xeon processors

[HEXUS]

And the Xeon E-2100 entry level server processors, with up to 6C/12T, start to ship.

Read more.

[philehidiot]

Anyone else getting sick of Intel treating their customer base for high end parts (i.e. their most knowledgeable and discerning customers) like idiots and trying to con them?

Hypothetical: I'm about to drop several hundred thousand on processors for a server room. Or I'm about to spend £2000 on a personal gaming system. Don't worry, Intel, I'll take your word for it and not even bother looking around for the best way to spend money. You just continue to lie to me and have it pointed out by the press. It won't affect my purchasing decision in any way.

Bunch of prats.

Comet did something very similar with a mains filter that supposedly made everything better. They had two TVs next to each other, one all washed out and fuzzy and one bright and clear. When I looked into it, they'd changed the settings on the TVs to make one utterly washed out and dim and one fine and bright and sharp and then, after I'd forced him to fix that, the "salesman" (read: con artist) pointed out one was still better than the other. A quick inspection showed one was running on SCART and the other HDMI.

Comet went bust a few weeks later. Trying to con your customers is a BAD MOVE.

[Andy00900]

yep, even the pricing on the consumer 9th series are a step in the wrong direction.

[LSG501]

LOL... rigged 'benchmarks' and basically ripping off threadrippers approach to cpu design just shows how shaken Intel are.

And in related news AMD announces their latest version of threadripper with 64 cores a few months early...
https://www.servethehome.com/amd-epyc-rome-details-trickle-out-64-cores-128-threads-per-socket/

Note that the last bit isn't real but I can see it happening lol

[Iota]

LOL... rigged 'benchmarks' and basically ripping off threadrippers approach to cpu design just shows how shaken Intel are.

And in related news AMD announces their latest version of threadripper with 64 cores a few months early...
https://www.servethehome.com/amd-epy...ds-per-socket/

Note that the last bit isn't real but I can see it happening lol

Interesting that we may see PCIe gen 4 introduced with Rome, maybe that means Zen2 will support it as well. God I'm enjoying AMD kicking Intel in the goolies.

[watercooled]

Oh the irony. Have they licensed AMD's 'glue' then?

[CAT-THE-FIFTH]

Oh the irony. Have they licensed AMD's 'glue' then?

This is an MCM?? Oh!!

[Terbinator]

Weren't the original Dual-Core from Intel (and AMD) MCMs anyway back in the 00s?

[CAT-THE-FIFTH]

Were the original Dual-Core from Intel (and AMD) MCMs anyway back in the 00s?

Not sure,but the first Core2 Quads were.

[GinoLatino]

Were the original Dual-Core from Intel (and AMD) MCMs anyway back in the 00s?

What is "00s" the beginning of time?

[CAT-THE-FIFTH]

What is "00s" the beginning of time?

I assume the year 2000 to 2009?

[simonpreston]

Were the original Dual-Core from Intel (and AMD) MCMs anyway back in the 00s?

I think the Pentium Ds were. Perhaps also the first Core 2 Quads (not sure about the later ones).

[azrael-]

I somehow missed the "World's 1st professional multi-chip processor*". I'm sure that tagline MUST be there somewhere...

*Using world class patented IntelliGlue technology

[Tabbykatze]

UPI glue is better than Infinity Glue but you can tell theyvecompletely cannibalised their parts because they've indicated that the chips they've used are for an up to octa socket setup. Whereas these new MCMs can only be used in a dual socket setup.

Basically they've moved the processors from having their own socket to being on the same substrate and instead of mobo UPI, the dies are interconnected directly. Because of this, it is obvious Intel had no answer to Epyc and has literally knee jerked this product out the door to keep an advantage. But oh man those TDPs aren't nice...

[DanceswithUnix]

I think the Pentium Ds were. Perhaps also the first Core 2 Quads (not sure about the later ones).

Pretty sure it was only the Pentium D that was a 2 chip MCM. That socket was designed for cheap dual socket boards as you basically just wired the CPUs in parallel and they fought for the bus when they needed it, so Intel didn't need any glue logic on their MCM. It was conceptually very ugly, but for the software of the day where you were far more likely to be running two different programs than something multi threaded it worked well enough. Heck, the P4 wasn't nice in any guise, so adding another layer of ugly wasn't going to matter.

But oh man those TDPs aren't nice...

Have you seen a TDP for this beast? I had a quick look, Intel seem amusingly quiet on the subject so I guess they aren't proud

[Tabbykatze]

Have you seen a TDP for this beast? I had a quick look, Intel seem amusingly quiet on the subject so I guess they aren't proud

I'm trying to see where i saw it but i could have sworn i saw 350w tdp. Considering the 8180 is 205w TDP and (205w/28)x48 gives 351, it's a reasonable assumption if it's not confirmed. The only way it could have a lower tdp is if they downclock it further than the 8180 but we also don't know the thermal impact of UPI running on chip rather than mobo.

Obviously multiplying out the per core tdp is not an accurate measure but its a good estimate while i try to find that article.